Abstract：A controllable mosaic algorithm based on single-scale Retinex was proposed for two partially overlapping finger vein
images from small image acquisition module. Firstly, finger vein images were enhanced by a modified single-scale Retinex.
Subsequently, the feature points from two finger vein images were extracted and descripted by SIFT algorithm. Rough matched
feature points were obtained by similarity measure with Euclidean distance. RANSAC algorithm was employed to eliminate
error matched points and the homography matrix between two finger vein images was estimated. Meanwhile, constraints of the
perspectively transformed image contour vertices were set to control the results of the image mosaic. Two finger vein images were
stitched together and a weighted average fusion algorithm was used for seam elimination. The result of finger vein images mosaic can
be controlled through constraints of the perspectively transformed image contour vertices. Experimental results demostrate that this
algorithm is controllable, and suitable for finger vein images mosaic.
